Enhancing Prediction of Chlorophyll-a Concentration with Feature Extraction using Higher-Order Partial Least Squares

Taewhi Lee, Jang-Ho Choi, Miyoung Jang, Jongho Won, Jiyong Kim

Abstract: Harmful algal blooms can cause significant negative impacts on the health of humans and other organisms. It is useful to predict chlorophyll-a concentration accurately for the forecast of algal blooms. While convolutional machine learning models are often used for such prediction, they may not fully consider the relationships among input features. We propose an approach to enhance the prediction by extracting latent features using higher-order partial least squares (HOPLS). The experimental results show that the feature extraction using HOPLS can significantly improve the prediction accuracy, especially in critical cases.

External IDs:dblp:conf/ictc/LeeCJWK20